Our Quest to Become AI-First and What We Learned

Six months ago, Amplitude set out on a mandate to become an AI-first organization. We saw how the technology landscape was shifting, and we raced headfirst into a wholesale shift.

Here’s some data about what this evolution has looked like in practice:

• Over 1,000 AI agents and automated workflows developed internally across GTM and corporate in 6 months
• 1 super-agent called Moda that orchestrates across SFD, Jira, Slack, Amplitude, and more
• 50+ AI application POCs completed
• 12 AI applications in production, including Outreach AI, Slack AI, ZoomInfo Copilot, Gemini, GPT, Claude, Glean AI, Workato, Momentum, Workday, NetSuite AI, and MCP
• 30 enterprise agents operating in Workato
• 900+ agents developed in Glean AI
• 27 company-wide enablement and training sessions
• 7 non-engineering AI hackathons

Our growth over the past 6 months hasn’t come from perfect technology. Most of it is far from perfect. It’s come as a result of focus, repetition, and a whole lot of education along the way. Here are the five most important things we’ve learned so far:

1. The AI-first approach is not about tools, it’s about execution

Becoming AI-first is often framed as a technology decision. I’ll tell you firsthand that it’s not. In practice, it’s all about execution.

At Amplitude, we focused on applying AI everywhere we could, but we largely agreed that:

1. Employees would continue to do the work
2. AI adoption would be uneven and messy
3. Meaningful ROI would take time and iteration

2. It’s still early for AI in GTM

The promise of AI in go-to-market workflows is real. The use cases are becoming increasingly clear. But it’s hard to change GTM workflows without a concrete understanding of how AI impacts day-to-day habits. The tools are evolving as we implement them, and the value isn’t always fully baked.

For us, success has meant staying grounded in what works today rather than what is theoretically possible. Focus on solving one problem at a time, and start small. Once you’ve mastered the basics, you can take bigger steps.

3. The shift that mattered most was from tools to workflows

Early experimentation focused on AI tools themselves. This resulted in some small wins for us, but didn’t really amount to much at scale. The real progress came when we shifted our attention from tools to specific workflows and use cases.

Instead of starting with the technology and working backward to fit it in, we started with the use cases and asked which AI tools and workflows could best serve our needs. We reduced friction immediately and improved outputs across the business faster than expected. This helped us stay focused on practical gains that actually made a difference to our employees.

Today, AI supports valuable work across the business, including:

• Account planning with AI-driven research and recommendations
• Sales agents for deal framing and talk tracks
• Post-sales transitions with customer call context
• Renewal workflows
• Customer call intelligence flowing into CRM and product teams
• Localized content and training generated for global teams

4. Scaling AI means making tradeoffs

As AI adoption expands, so does the complexity. At Amplitude, we found that point solutions move quickly while platforms scale better. Meanwhile, some issues that we knew we’d run into (e.g., Agent observability and security concerns) surfaced earlier than expected. If your team plans to benefit from AI adoption in the short or long term, you need to also make a plan to manage the tradeoffs you're expecting.

We also found that consistency is what matters most once AI touches GTM execution. As a result, we’ve focused less on picking the right tools and more on managing tradeoffs deliberately.

5. Structure and enablement matter

One of the most important enablers for us was establishing a formal internal AI program. This included clear leadership, dedicated enablement, and technical support. We also established an AI Center of Excellence responsible for tool evaluation, proof of concepts, governance, and security.

Just as important, this work was prioritized over competing internal projects. AI was treated as a core initiative for the CIO organization rather than just another side project.

The takeaway

Becoming AI-first didn’t mean driving faster results everywhere or looking for immediate ROI. It didn’t start with the most exciting AI technology. Instead, we began by changing employee habits and adapting new tools to the reality of existing workflows. We embraced experimentation, invested in enablement, and gave teams time to learn.

At the end of the day, AI does not create leverage on its own. People and execution do.